Journal of Software - Academy Publisher
Journal of Software - Academy Publisher
Journal of Software - Academy Publisher
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
844 JOURNAL OF SOFTWARE, VOL. 6, NO. 5, MAY 2011<br />
completely, quote the case directly and make a<br />
conclusion.<br />
(b) Modifying<br />
It takes fault phenomenon vector as input and to reason<br />
with explained distance discrimination rules and past<br />
fault data. The reasoning machine issued a diagnostic<br />
probability order table <strong>of</strong> diagnosis result according to<br />
discrimination analysis rules, where the fault with<br />
maximum probability is the preferred result, and others<br />
are options by decreasing order <strong>of</strong> probability. If the case<br />
is not matched completely, The diagnostic probability<br />
order table will be available to maintenance personnel for<br />
reference <strong>of</strong> further confirm, use the Distance<br />
determination rule database, parts fault characteristic and<br />
actions record [16] etc. to Reasoning, adjust, rewrite,<br />
match and synthesize the case which has been retrieved<br />
according to the current fault phenomenon <strong>of</strong> equipment.<br />
(c) Storing<br />
Make the corrected case in keeping with the diagnosis<br />
<strong>of</strong> the current fault phenomenon, and make a conclusion.<br />
At the same time, the confirmed result will be fed back to<br />
fault database for record to prepare for the next diagnostic<br />
reference.<br />
The distance discrimination was used to fault diagnosis<br />
to establish expert system based on fault phenomenon<br />
vector. The core <strong>of</strong> fault diagnosis is that it can<br />
memorize/store the former fault, its environments and the<br />
process accurately, furthermore, it uses the past diagnosis<br />
experience, process and methods to complete the current<br />
diagnosis through analogy and association while<br />
diagnosing. Therefore, fault diagnosis based on fault<br />
phenomenon vector is a kind <strong>of</strong> methods realized through<br />
analogy [17, 18], and its design mode is to utilize the past<br />
designed case directly instead <strong>of</strong> the summary <strong>of</strong> design<br />
experience.<br />
III. KEY TECHNIQUES<br />
A. Rule-Based Diagnostic Expert Systems<br />
In the rule-based systems, knowledge is represented in<br />
the form <strong>of</strong> production rules. A rule describes the action<br />
that should be taken if a symptom is observed. The<br />
empirical association between premises and conclusions<br />
in the knowledge base is their main characteristic. These<br />
associations describe cause-effect relationships to<br />
determine logical event chains that were used to represent<br />
the propagation <strong>of</strong> complex phenomena. The general<br />
architecture <strong>of</strong> these systems includes domain<br />
independent components such as the rule representation,<br />
the inference engine and the explanation system. Basic<br />
structure <strong>of</strong> a classical rule-based expert system is shown<br />
in Fig. 2.<br />
Expert diagnosis experiences suitably formatted<br />
consists the basis for the classical expert system approach.<br />
Fault diagnosis requires domain specific knowledge<br />
formatted in a suitable knowledge representation scheme<br />
and an appropriate interface for the human-computer<br />
dialogue. In this system the possible symptoms <strong>of</strong> faults<br />
are presented to the user in a screen where the user can<br />
click the specific symptom in order to start a searching<br />
© 2011 ACADEMY PUBLISHER<br />
Fault<br />
database<br />
Knowledge<br />
base<br />
Distance<br />
determination rule<br />
database<br />
User<br />
interface<br />
maintenance<br />
personnel<br />
Inference<br />
engine<br />
Expert<br />
Figure 2. Basic structure <strong>of</strong> a rule-based expert system.<br />
process for the cause <strong>of</strong> the fault. Additional information<br />
about checking or measurements is used as input that, in<br />
combination with stored knowledge in the knowledge<br />
base guide to a conclusion [19, 20, 21, 23].<br />
B. Reasoning Rules Formulation<br />
The formulation <strong>of</strong> rules needs to resolve problem <strong>of</strong><br />
fault data table design. Table 1 is the designed fault data<br />
table <strong>of</strong> F1, where each line represents a fault<br />
phenomenon vector.<br />
Using the above method, we can build fault data table<br />
for each Fi. Each fault phenomenon obeys standard 0-1<br />
distribution, the value <strong>of</strong> which is shown in (1). The<br />
expectation <strong>of</strong> each phenomenon is pij, where i represent<br />
that the phenomenon is caused by the i-th fault; j<br />
F1<br />
TABLE I.<br />
DATA TABLE OF F1<br />
Phenomenon<br />
Number I1 I2 I3 I4 … Im<br />
1 1 0 0 1 … 1<br />
2 1 0 1 0 … 0<br />
3 0 0 1 1 … 0<br />
4 1 0 1 0 … 0<br />
5 1 0 1 1 … 0<br />
6 1 0 1 0 … 0<br />
7 1 0 1 1 … 0<br />
8 0 0 0 1 … 0<br />
9 1 0 0 0 … 0<br />
10 1 1 1 0 … 0<br />
… … … … … … …<br />
Total 1000 913 11 946 583 … 50